Researchers at UC Berkeley have created an AI robot capable of folding laundry in record time. Check out the video demonstration, and see how this invention could simplify everyday life.
Folding laundry is no picnic. Thanks to researchers at UC Berkeley’s AUTOLAB, this chore may soon be a thing of the past.
The team has developed a robotic method for folding clothes called SpeedFoldingand has just broken the speed record. Thanks to computer vision technology, a neural network called BiManual Manipulation Network (BiMaMa-Net), and a pair of industrial robot armsthe system can fold 30 to 40 randomly arranged garments in an hour.
You need about two minutes to SpeedFolding to fold a garment. This may seem a long time by human standards, but previous methods only achieved 3-6 folds per hour. The record has been broken.
One of the first two-arm folding robots
As scientists, the researchers have tackled the problem of linen folding in a very technical way. According to them, “ folding garments reliably and efficiently has long been a challenge for robotic handling, due to the complex dynamics and high-dimensional configuration space of garments. “.
To put it simply, clothes wrinkle easily and can be strewn all over the room. That’s why researchers have created a two-arm system capable of ironing a garment and then folding it.
Previous robotic methods of folding laundry mainly used a single arm. This explains the enormous increase in speed offered by SpeedFolding.
A neural network trained on 4000 examples
In addition, the BiMaMa-Net neural network has learned to fold laundry. after studying 4300 examples. These included both human demonstrations and machine-assisted bending.
Now, the industrial robot associated with this AI can fold garments from a random initial position in less than 120 seconds with a 93% success rate.
In addition, SpeedFolding can apply what it learns to garments from different materials, shapes or colors than the clothes on which he was trained.
The system begins by examining the initial state of the garment, using a camera. He then calculates how to grab it with both arms, to reach the next stage of the folding process.
The price of not having to deal with laundry
The industrial robot used for this project is an ABB YuMi equipped with grippersextended by a small 3D-printed tooth to improve grip. A machine costs around $58,000.
Consequently, this robotic system is unlikely to find its way into every home. We’ll just have to carry on doing our chores for a few more years…
However, in the long term, this technique could find its place in home robotics. Soon, robots will be marketed to the general public and will be able to apply this method.
For the time being, as potential applications for this technology, the team of scientists mentions textile manufacturing, logistics, industrial laundries, healthcare and hospitality.
Visit researchers involved in this project are Yahav Avigal, Tamim Asfour, Torsten Kröger and Ken Goldberg from UC Berkeley, and Lars Berscheid from the Karlsruhe Institute of Technology. They will be presenting SpeedFolding at IROS2022, to be held next week in Kyoto.
What about you? Are you interested in domestic robots? What are the possible applications for such technology?